An Artificial Immune Algorithm For Dynamic Task Scheduling On Cloud Compution Platform

From grid computing, distributed computing, the algorithm of task scheduling has become the focus of academic and industry research. Cloud computing must achieve two important goals: scalability and high availability cloud computing, as well as the ability to deal with data intensive and computation intensive, communication intensive and other types of services. Meanwhile, cloud computing needs higher requirement to the system, such as energy conservation, efficiency, stability and reliability. The reasonable task scheduling is one of the ways to realize these two important goals of cloud computing.In addition, as an indispensable part of cloud computing, the virtualization technology should also be combined with task scheduling algorithm. Many algorithms on task scheduling now can not complete the task in new technology requirement. The theory and method from artificial intelligence(such as genetic algorithm) is always one of important methods when task scheduling is under complex computing environment. Then, artificial immune system is one of focus problems in artificial intelligence, and it has many advantages than genetic algorithm. In this thesis, we will introduce the artificial immune system into the dynamic task scheduling problems on cloud computing platform.We propose an Artificial Immune Algorithm for Dynamic Task Scheduling on Cloud Computing Platform in this paper. Firstly, the algorithm uses the queuing theory to determine the conditions of cloud computing platform to maintain steady-state. Such as, the minimum services ratio is needed under different tasks in this platform. Then, the dynamic task scheduling problem is modeled as an allocation of computing resource for virtual machines, based on the immune clone selection algorithm in artificial immune system. Based on the CPU processing power for all kinds of tasks, we designed and implemented a stochastic search algorithm for virtual machine resource allocation scheme. Finally, proper load balancing processing is put into the immune system, which provides more excellent antibody genes.The results of simulation experiment show that, compared with hybrid genetic scheduling algorithms, this algorithm for dynamic task scheduling on cloud computing platform which based on artificial immune, can effectively improve the convergence speed and accuracy, is fast to get the reasonable allocation, and improve the cluster resource utilization.